Analysis of lung cancer by machine learning techniques

In this study, we aimed to better diagnosis of lung cancer with lung tomography, lung cell histopathology and light microscopy images. In addition, two types of images, good and malignant, were used. Various size reduction methods have been used to obtain optimum properties of these two types of images. These methods include Principal Component Analysis (PCA), Generalized Discriminant Analysis (GDA), Linear Discriminant Analysis (LDA), Local Linear Embedding (LLE), Classical Multidimensional Scaling (MDS), Neighborhood Preserving Embedding (NPE), Stochastic Proximity Embedding (SPE) ) methods. Naive Bayes (NB), Random Forest (RF), Decision Tree (DT) methods were used as the classifier. In addition, Maximum Stationary Extreme Regions (MSER), Speeded Up Robust Features (SURF), Histogram of Orianted Gradients (HOG) were added to each of these classifier methods and the status of the accuracy value was examined. Various methods have been developed for the early and accurate diagnosis of lung cancer. In this study, a total of 240 images were used for the methods. These images were obtained from 60 patients by taking 4 sections for each. The accuracy of each classifier, each size reduction method and each feature was compared with the images used. In addition, the methods with the highest accuracy were found. Some of the methods used were found to have 100% accuracy. Thus, the distinction between benign and malignant tumors has been achieved thanks to the methods that provide the highest accuracy. It is also aimed to increase the accuracy rate for the early diagnosis of lung cancer, which is one of the cancer types with the highest mortality rate in the world. Thus, lung images loaded into computer assisted diagnostic systems were passed through various stages in this system to obtain an accurate result. Key Words: Lung Cancer, Machine Learning, Classification, Image Processing, Histopathology, Lung Tomography, Computer Assisted Diagnosis, Computed Tomography.